High Water Mark
In the mainframe context, a **High Water Mark** refers to the maximum value or peak level that a specific resource, metric, or counter has reached over a defined period. It indicates the highest point of utilization or consumption observed, providing critical insight into peak demand or resource requirements.
Key Characteristics
-
- Resource Tracking: Commonly tracks resources such as storage (e.g.,
GETMAINrequests, buffer pool usage), CPU utilization, dataset size, queue depth (e.g., MQSeries, CICS temporary storage), or the number of active tasks. - Peak Value Indicator: Represents the absolute maximum observed value, not an average, providing a critical data point for understanding system stress or capacity limits.
- Reset Mechanism: High water marks are often reset, either manually by an operator, automatically at specific intervals (e.g., system IPL, daily, weekly), or upon certain events, allowing for tracking of new peak periods.
- Monitoring and Reporting: Typically captured and reported by system monitoring tools (e.g., RMF, OMEGAMON), subsystem utilities (e.g., DB2, CICS statistics), or specific application logic.
- Persistence: Can be transient (cleared on system restart) or persistent (stored in logs or control blocks) depending on the resource and monitoring implementation.
- Resource Tracking: Commonly tracks resources such as storage (e.g.,
Use Cases
-
- Capacity Planning: Analyzing high water marks for CPU, memory, and I/O helps determine future hardware or software resource requirements to accommodate peak workloads.
- Performance Tuning: Identifying peak buffer pool usage (e.g., DB2, IMS) via high water marks can guide adjustments to buffer pool sizes to prevent I/O bottlenecks or storage constraints.
- Problem Determination: Unexpectedly high water marks for certain resources (e.g., CICS storage, MQ queue depth) can signal a runaway task, a loop, or an application error requiring investigation.
- Resource Allocation Optimization: Using high water marks for dataset sizes can help optimize space allocation, preventing
X37abends while avoiding overallocation. - Workload Management (WLM) Tuning: Understanding peak resource demands helps in defining appropriate WLM service classes and goals to ensure critical workloads meet their performance objectives.
Related Concepts
High Water Marks are intrinsically linked to performance monitoring and capacity management on z/OS. They are often reported by system measurement facilities like SMF (System Management Facilities) and RMF (Resource Measurement Facility), which collect detailed performance data. Subsystems like DB2, CICS, and IMS also maintain and report high water marks for their internal resources (e.g., buffer pools, storage areas, transaction counts) through their respective statistics or monitoring interfaces. They provide crucial input for Workload Management (WLM) decisions and storage management strategies, helping to ensure system stability and efficiency.
- Regular Monitoring: Routinely review high water marks for critical system resources and key applications to identify trends and potential bottlenecks before they impact performance.
- Establish Baselines: Define normal high water mark ranges for different workloads and time periods to quickly identify anomalies or unexpected spikes.
- Automated Alerting: Implement automated alerts for high water marks that exceed predefined thresholds, enabling proactive intervention for potential resource exhaustion or performance degradation.
- Capacity Planning Integration: Incorporate high water mark data into your capacity planning processes to make informed decisions about resource provisioning and upgrades.
- Historical Analysis: Maintain historical high water mark data to analyze long-term trends, anticipate growth, and validate the effectiveness of performance tuning efforts.